Let $A$ be a finite dimensional $\mathrm{C}^\ast$-algebra.
Let $I$ be a left ideal in $A$. I believe there is a projection $p\in A$ (an element such that $p=p^2=p^*$) such that:
$$I=Ap.\qquad(1)$$
Is there an easy-to-write down proof, or easy reference for this?
I have:
As a (norm-)closed left ideal of a finite dimensional $\mathrm{C}^\ast$-algebra, $I$ is a weakly closed left ideal of a von Neumann algebra, and thus $I$ must be of the form $F(G)p$ for a projection $p$.
This is obvious overkill (or am I actually mistaken about (1)?)
Here some steps to get $p$.
$I$ contains a nonzero projection. Indeed, if $I\ne\{0\}$, take $a\in I$ nonzero. Then $a^*a\in I$. Then $f(a^*a)\in I$ for all $f\in\mathbb C[x]$. Because $A$ is finite-dimensional, any spectral projection of $A$ can be obtained as $f(a^*a)$.
$I$ contains a nonzero minimal projection. Indeed, given a projection $q\in I$, let $q_0\leq q$ be a minimal projection in $A$ (which exists because $A$ is finite-dimensional). Then $q_0=q_0 q\in I$.
There exist minimal projections $q_1,\ldots,q_m\in I$ such that $I=A(q_1+\cdots+q_m)$. Indeed, let $q_1\in I$ be a nonzero minimal projection. Consider the left ideal $I(1-q_1)$. Note that $I(1-q_1)\subset I$: if $b\in I$ then $b-bq_1\in I$. The steps above apply to $I(1-q_1)$, if nonzero, so there exists a minimal projection $q_2\in I(1-q_1)\subset I$. Then $q_2(1-q_1)=q_2$, which gives $q_2q_1=0$. We can repeat this process with $q_1+q_2$, etc., until at some point $I(1-q_1-\cdots-q_m)=\{0\}$.
Take $p=q_1+\cdots+q_m\in I$. We have $I=Ip$. Then $Ip\subset Ap\subset I=Ip$.